Chapter 15: Principal Component Analysis
Dimension reduction is one of the more important concepts/strategies in machine learning. It is sometimes equated with feature selection, but that is too narrow a view of dimension reduction. Our models often have to deal with an excess of features, some of which are capturing the same information. Not addressing the issue substantially increases the risk of overfitting or of unstable results. But dropping some of our features is not the only tool in our toolbox here. Feature extraction strategies, such as principal component analysis (PCA), can often yield good results.
We can use PCA to reduce the dimensions (the number of features) of our dataset without losing significant predictive power. The number of principal components necessary to capture most of the variance in the data is typically less than the number of features, often much less.
These components can be used in our regression or classification models rather than the initial...
